[MRG] Faster gromov-wasserstein linesearch for symmetric matrices

[KrzakalaPaul]

Types of changes

Add a new argument "symmetric" to ot.gromov._gw.solve_gromov_linesearch. The default value symmetric=False results in the previous behaviour. When C1 and C2 are symmetric however, one can set symmetric=True which results in faster computations as the formula can be simplified in that case.

ot.gromov._gw.fused_gromov_wasserstein and ot.gromov._gw.gromov_wasserstein have been changed accordingly to automatically pass symmetric=False or symmetric=True to solve_gromov_linesearch.

A new file symmetric_linesearch.py has been added to examples.gromov to check the usefullness of setting symmetric=True.

Motivation and context / Related issue

The formula for solving the gromov linesearch can be simplified in the (very common case) where the matrices C1 and C2 are symmetric, resulting in a small (but free) improvement in computationnal cost.

How has this been tested (if it applies)

PR checklist

• I have read the CONTRIBUTING document.
• The documentation is up-to-date with the changes I made (check build artifacts).
• All tests passed, and additional code has been covered with new tests.
• I have added the PR and Issue fix to the RELEASES.md file.

[codecov]

Codecov Report

Merging #607 (5e28029) into master (737b20d) will increase coverage by 0.00%.
The diff coverage is 100.00%.

Additional details and impacted files

@@ Coverage Diff @@ ## master #607 +/- ## ======================================= Coverage 96.78% 96.78% ======================================= Files 77 77 Lines 16002 16004 +2 ======================================= + Hits 15488 15490 +2  Misses 514 514 

[cedricvincentcuaz]

Good catch @KrzakalaPaul , thank you :)
Could you edit the releases.md with your PR ? then the PR will be ready for merging.

Best, Cédric.

[KrzakalaPaul]

Hi :)
I tried to edit releases.md, tell me if it's alright.
Best, Paul